Menu
SkillsMP has collected 31 skills from GregoryHo/arcforge. Open a skill to review its source and details.
Showing 31 of 31 collected skills.
Use this skill whenever the user (an arcforge contributor) says they want to bump arcforge's version, cut a release, "ship vX.Y.Z", "準備發版", "ready to release", or any equivalent intent on the arcforge repo itself — even if they don't use the word "release".…
Execute a task list with one fresh subagent per task plus a single task-reviewer each. Use when tasks need isolated contexts to avoid cross-task pollution; for in-session human-checkpoint execution use arc-executing-tasks.
Read-only advisory audit of an SDD spec family — design, spec, dag, and decision anchors.
Explore and shape a design before implementation. Use when intent is vague, a new feature needs design, or an existing spec needs another iteration — the front door to the SDD pipeline that writes design.md.
Guide manual context compaction to workflow phase boundaries. Use when deciding whether to compact now — compact at clean seams between phases rather than mid-task; the compact-suggester hook may route here as context grows.
Coordinate multi-epic worktrees and cross-epic DAG state. Use when specs/<id>/dag.yaml exists and epics run in parallel worktrees needing sync and merge orchestration; to build a single epic's features use arc-implementing.
Investigate root cause systematically before proposing any fix. Use when hitting a bug, test failure, or unexpected behavior — reproduce, isolate, and confirm the cause with evidence rather than guessing at patches.
Create an Excalidraw diagram or visual — architecture, flowchart, mind map — including casual 'draw this' / 'show me visually'. Also use when arc-maintaining-obsidian delegates Synthesis visuals beyond embedded Mermaid.
Dispatch multiple independent features to parallel subagents within one worktree session. Use when fanning out feature-level work you drive yourself; for epic-level teammates you monitor as a present lead use arc-dispatching-teammates.
Run epic-level parallel work via agent teammates while staying present to monitor. Use when dag.yaml has 2+ ready epics and you want a live team; for walk-away loops use arc-looping, for feature fan-out use arc-dispatching-parallel.
Measure whether a skill, agent, or workflow actually changes AI agent behavior. Use when shipping a new skill, after modifying one, or comparing alternative approaches — the eval gate behind the Iron Law.
Run a prepared task list yourself with human-in-the-loop checkpoints. Use when tasks are already broken down and you implement them in-session; for isolated subagent-per-task execution use arc-agent-driven instead.
Integrate finished work once implementation is complete and tests pass. Use when deciding how to merge — Step 0 discriminates epic worktrees (.arcforge-epic present) from regular branches and runs the matching path.
Orchestrate implementation of a large multi-feature project inside a worktree. Use when driving one epic's build end-to-end; for cross-epic worktree and DAG-state management use arc-coordinating instead.
Capture session reflections into a durable diary. Use when a significant work session ends, the user asks to journal, or the PreCompact hook fires before context is compacted and insight would be lost.
Run the opt-in learning lifecycle: turn observations into reviewable candidates, inactive drafts, then explicitly activated instincts. Use when the default-off learning module is enabled and observations should become artifacts.
Run an autonomous unattended loop that executes DAG tasks across sessions without human intervention. Use when walk-away execution is needed; for a present lead monitoring epic teammates use arc-dispatching-teammates instead.
Manage an Obsidian vault: ingest, query, audit, or initialize. Use when saving notes, ideas, or URLs ('save this' / 'file this back'), querying vault knowledge, or auditing health (orphans, drift). Not for diagrams — use arc-diagramming-obsidian.
Hand off, save, or resume session state across turns. Use when ending a session and handing to a future one, summarizing recent context to continue, or archiving, resuming, or aliasing a saved session for durable reference.
Break a refined spec into an executable DAG of epics. Use when specs/<id>/spec.xml exists and needs decomposition into dependency-ordered epics and features before implementation — produces dag.yaml for arc-coordinating.
Manually save a pattern or insight from the current session as a reusable instinct.
Formalize an approved design into a structured spec.xml with acceptance criteria. Use when a design.md is ready to become the authoritative spec, or spec quality is below threshold — the delta authority, hands to arc-planning.
Analyze accumulated diary entries for recurring patterns and preferences. Use when 5+ diaries have accumulated, the user asks to summarize learnings from past sessions, or inject-context flags that reflection is due.
Optimize a measurable metric through autonomous hypothesis-driven experimentation. Use when a target has a numeric signal — build times, algorithm efficiency, prompt quality, model performance — and you want iterative measured gains.
Request code review then process the returning feedback with technical rigor. Use when a task or feature is complete and needs review — runs the full request-to-receive loop, dispatching the code-reviewer agent and handling its verdict.
Drive implementation test-first through RED, GREEN, REFACTOR. Use when about to write any feature or bugfix code — write the failing test first, then the minimum code to pass, then refactor. Enforces the Iron Law for implementation.
Route an ArcForge task to the smallest useful skill or workflow. Use when unsure which skill applies or the user asks where to start — a bounded router and skill index, not an always-on policy engine.
Create an isolated git worktree in ANY repo — a parallel branch, experiment, or review checkout. Use when work needs isolation even if the user never says 'worktree'; epic context auto-escalates to arc-coordinating, else the generic CLI.
Gather fresh evidence that work is actually complete before a completion claim. Use when about to say 'done', 'fixed', or 'passing' — rerun the checks and read real output rather than trusting prior state.
Create, edit, or verify ArcForge's own skills and skill tests before deployment.
Break a feature into small executable tasks with exact code and commands. Use when a spec or feature is ready and you need concrete implementation steps before coding — produces the task list arc-executing-tasks runs.